diff options
author | Michel Iwaniec <michel.iwaniec@arm.com> | 2017-11-29 10:48:23 +0000 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:42:17 +0000 |
commit | 5dfeae62f89eefdc241887c3e67cd1c04ec0b6a7 (patch) | |
tree | d6b5d40353aa68aeda803c809812fd6e208c3e7f /src/core/NEON | |
parent | 7f0f790ae7f5dd044a5d7564492583b8df974a11 (diff) | |
download | ComputeLibrary-5dfeae62f89eefdc241887c3e67cd1c04ec0b6a7.tar.gz |
IVGCVSW-820: Add QASYMM8 support to NeonActivationLayerKernel
Change-Id: Ic3881e97b4fcbae0ac287a1e010cfc6f0fd8d7d1
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/112139
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'src/core/NEON')
-rw-r--r-- | src/core/NEON/kernels/NEActivationLayerKernel.cpp | 56 |
1 files changed, 55 insertions, 1 deletions
diff --git a/src/core/NEON/kernels/NEActivationLayerKernel.cpp b/src/core/NEON/kernels/NEActivationLayerKernel.cpp index 6ea504a173..9670b7798c 100644 --- a/src/core/NEON/kernels/NEActivationLayerKernel.cpp +++ b/src/core/NEON/kernels/NEActivationLayerKernel.cpp @@ -26,8 +26,10 @@ #include "arm_compute/core/FixedPoint.h" #include "arm_compute/core/Helpers.h" #include "arm_compute/core/ITensor.h" +#include "arm_compute/core/NEON/NEAsymm.h" #include "arm_compute/core/NEON/NEFixedPoint.h" #include "arm_compute/core/NEON/NEMath.h" +#include "arm_compute/core/QAsymm8.h" #include "arm_compute/core/TensorInfo.h" #include "arm_compute/core/Utils.h" #include "arm_compute/core/Validate.h" @@ -44,7 +46,7 @@ namespace Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::QASYMM8, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32); // Checks performed when output is configured if((output != nullptr) && (output->total_size() != 0)) @@ -107,6 +109,9 @@ void NEActivationLayerKernel::configure(ITensor *input, ITensor *output, Activat ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), (output != nullptr) ? output->info() : nullptr)); + ARM_COMPUTE_ERROR_ON_MSG((input->info()->data_type() == DataType::QASYMM8) && (activation_info.activation() != ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU), + "For QASYMM8 only lower/upper bounded relu is supported"); + // Activation functions : FP32 static std::map<ActivationFunction, ActivationFunctionExecutorPtr> act_map_f32 = { @@ -170,9 +175,17 @@ void NEActivationLayerKernel::configure(ITensor *input, ITensor *output, Activat { ActivationFunction::SQUARE, &NEActivationLayerKernel::activation<ActivationFunction::SQUARE, qint16_t> }, { ActivationFunction::TANH, &NEActivationLayerKernel::activation<ActivationFunction::TANH, qint16_t> }, }; + // Activation functions : QASYMM8 + static std::map<ActivationFunction, ActivationFunctionExecutorPtr> act_map_qasymm8 = + { + { ActivationFunction::LU_BOUNDED_RELU, &NEActivationLayerKernel::activation<ActivationFunction::LU_BOUNDED_RELU, qasymm8_t> }, + }; switch(input->info()->data_type()) { + case DataType::QASYMM8: + _func = act_map_qasymm8[activation_info.activation()]; + break; case DataType::QS8: _func = act_map_qs8[activation_info.activation()]; break; @@ -546,6 +559,47 @@ typename std::enable_if<std::is_same<T, int8_t>::value, void>::type NEActivation } template <ActivationLayerInfo::ActivationFunction F, typename T> +typename std::enable_if<std::is_same<T, qasymm8_t>::value, void>::type NEActivationLayerKernel::activation(const Window &window) +{ + Iterator input(_input, window); + Iterator output(_output, window); + const QuantizationInfo qi_in = _input->info()->quantization_info(); + const QuantizationInfo qi_out = _output->info()->quantization_info(); + const qasymm8x16_t a = vdupq_n_u8(sqcvt_qasymm8_f32(_act_info.a(), qi_in.scale, qi_in.offset)); + const qasymm8x16_t b = vdupq_n_u8(sqcvt_qasymm8_f32(_act_info.b(), qi_in.scale, qi_in.offset)); + // Initialise scale/offset for re-quantization + float s = qi_in.scale / qi_out.scale; + float o = -qi_in.offset * s + qi_out.offset; + float32x4_t vs = vdupq_n_f32(s); + float32x4_t vo = vdupq_n_f32(o); + + execute_window_loop(window, [&](const Coordinates & id) + { + const auto input_ptr = reinterpret_cast<const qasymm8_t *>(input.ptr()); + const auto output_ptr = reinterpret_cast<qasymm8_t *>(output.ptr()); + + const qasymm8x16_t in = vld1q_u8(input_ptr); + qasymm8x16_t tmp = {}; + + switch(F) + { + case ActivationFunction::LU_BOUNDED_RELU: + // Perform activation + tmp = vminq_u8(a, vmaxq_u8(b, in)); + // Re-quantize to new output space + tmp = vmlaq_qasymm8(tmp, vs, vo); + break; + default: + ARM_COMPUTE_ERROR("Function not implemented"); + break; + } + + vst1q_u8(output_ptr, tmp); + }, + input, output); +} + +template <ActivationLayerInfo::ActivationFunction F, typename T> typename std::enable_if<std::is_same<T, qint16_t>::value, void>::type NEActivationLayerKernel::activation(const Window &window) { Iterator input(_input, window); |